What counts as proof, and what does not

Spend an hour looking at Telegram development studios in this region and a pattern emerges. Grids of client logos with no link to anything. Testimonials attributed to a first name and an industry. Percentages — forty per cent more bookings, sixty per cent fewer support tickets — with no methodology, no baseline and no way to ask the client whether they recognise the figure. Award badges from organisations that sell the badges.

Almost none of it is checkable, and a meaningful share of it is invented outright, which is why the whole format has stopped working. A buyer who has been shown five identical logo walls in a week does not read the sixth as evidence. They read it as the thing everyone puts on the page.

There is also a legal dimension here that is specific to the UAE. Consumer protection law treats misleading commercial statements as an offence rather than as marketing enthusiasm, and a fabricated testimonial or an invented certification is exactly that. Publishing one is not a grey area with an upside; it is exposure with a downside and no corresponding benefit, because the buyers who matter did not believe it anyway.

So the policy on this site is narrow and absolute: nothing appears as proof unless a reader can go and confirm it without asking us for anything. In practice that means two kinds of thing — a bot that is running, and code that is readable.

A live bot qualifies. If a link opens a conversation in Telegram and the thing responds, you have verified more in ten seconds than a case study could tell you in a page. You can see how it handles a wrong answer, what it does with an unexpected message, and whether the copy reads like someone thought about it.

A public repository qualifies. Code is the least ambiguous evidence available in this trade: it shows whether idempotency was handled, whether errors are caught or swallowed, whether the tests exist. It is also the evidence most studios will not offer, which tells you something on its own.

A named client qualifies only with written consent, and any number attached to them qualifies only if they have approved that exact figure. Both conditions are frequently impossible — plenty of businesses do not want it known that their bookings run through a bot — and an entry without a client name is not a weaker entry. An anonymous bot you can open beats a named logo you cannot.

What does not qualify, in any circumstance: a metric we calculated ourselves, a quotation we wrote on a client's behalf, a badge we bought, a client name used without permission, or a screenshot of something that is no longer running.

The reference material on this site is itself a form of evidence, and it is the kind that is hard to fake. The glossary documents the Bot API limits that actually bite — message rates, file sizes, payload caps, the forty-eight hour deletion window — with the exact numbers and working code. Anyone can check those against Telegram's own documentation, and anyone who has built against the API can tell within a paragraph whether the person writing knew what they were describing.

The same applies to the integration pages, which lead with the constraint rather than the capability: which providers will not underwrite a virtual asset business, which will invoice for a category their policy forbids, and the specific ways each one fails in production. That information comes from having hit those walls, and it is not available from a marketing site that has not.

The comparison pages carry a section listing the cases where the alternative is the right answer and we are not. Several of them recommend a forty-dollar-a-month flow builder outright. A studio inventing its credentials does not publish the argument against hiring it.

As live bots and public repositories become available to link, they appear here with the link attached. Until they do, this page carries the policy rather than a placeholder, because a portfolio section populated with plausible-looking examples is precisely the failure this whole approach exists to avoid.

If you are evaluating this studio and want more than a policy statement, the fastest route is a conversation in the bot. We will talk through a specific problem you have — the failure mode, the constraint you are going to hit, what it will cost — and you can judge the answer. That is a harder test to pass than a logo grid and a considerably more honest one.